<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Aditya Dhir on Medium]]></title>
        <description><![CDATA[Stories by Aditya Dhir on Medium]]></description>
        <link>https://medium.com/@adityadhir97?source=rss-34c84bf846ce------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*XrXpBF7ARE7ZAJTPptynXg.png</url>
            <title>Stories by Aditya Dhir on Medium</title>
            <link>https://medium.com/@adityadhir97?source=rss-34c84bf846ce------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 06:22:31 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@adityadhir97/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Part 1: MCP for Product Managers: You Were Always the Integration Layer]]></title>
            <link>https://medium.com/@adityadhir97/part-1-mcp-for-product-managers-you-were-always-the-integration-layer-0a29445370c2?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/0a29445370c2</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[product-management]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Fri, 05 Jun 2026 13:58:50 GMT</pubDate>
            <atom:updated>2026-06-05T14:05:52.920Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VwwpAjJNxYohaqqwARPljQ.png" /></figure><p>You might be the last PM in your company.</p><p>Not because AI is taking over product managers — that headline is lazy, and you’ve probably read it quite a few times by now. It’s because one specific version of this job is already dead, and sadly, many PMs haven’t noticed — they’re still doing the same things every day.</p><p>This is a guide to MCP for product managers — the Model Context Protocol — but not the kind you’ve already skimmed multiple times. I’m not writing it as a PM who wired up a couple of connectors over a weekend. I’m an engineer who built the plumbing underneath them.</p><p>Here’s the version of the job that’s becoming obsolete. For many years, in almost any company, the most expensive integration wasn’t software. It was you. Yes, you heard it right. As a PM, you’ve probably spent a significant chunk of your time pulling numbers out of PostHog, holding them in your notebook, retyping them into a PRD in Notion, and then breaking that PRD into tickets in Jira by hand. You were the bridge between systems that were never built to talk to each other. You were the integration layer — and that’s what MCP dissolves.</p><p>In the next sections, we’ll cover three things nobody who hasn’t built MCP will tell you: why your MCP servers keep silently disconnecting, why adding more tools makes your AI noticeably dumber, and the single piece of infrastructure that quietly fixes both. We’ll get to all three.</p><p>So if you’re thinking this is going to be another MCP explainer — well, you can find those anywhere. This is about MCP’s role in your job and which half of that job is worth keeping.</p><h3>You were always the integration layer</h3><p>Let’s talk about the last time you wrote a PRD.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bfF5W_GHPcQhxMHZ7RqqVA.png" /><figcaption>MCP for product managers diagram: before, a PM manually bridges PostHog, Notion, Jira and Slack (the M×N problem); after, MCP connects the tools to one AI assistant through a single protocol.</figcaption></figure><p>You opened PostHog and fetched the retention curve. You went through last month’s user interviews in Notion. You checked the existing bugs in Jira to see what was already known. You skimmed a Slack thread where support had flagged the same complaint three times. Then you sat down and started compiling all of it — translating five disconnected systems into one coherent document.</p><p>Phew. All of that felt like the job, but it wasn’t. It was a tax on the job. Each PM tool speaks a different language: PostHog thinks in events, Jira in tickets, Notion in pages. Each one is an isolated island, completely disconnected from the others. So you had to sit in the middle, read everything off them, and reconcile it. For the last couple of decades, that’s been a big part of the PM job — you act as a human API between systems that don’t integrate.</p><p>Engineers actually have a name for this problem. When you have M tools and N tasks that each need to touch several of those tools, you don’t get M-plus-N connections — you get M-times-N. Every tool times every workflow, each a separate bridge to build and maintain. It’s the integration problem that has haunted software for decades, and it’s exactly why your stack never connects the way you expect it to.</p><p>In most product organisations, the PM is how the company solves M×N — and no one has ever told you that out loud. It’s not the platform team, not the integrations team. It’s you. Every time you track down a metric and drop it into a document, you’re closing one of those M×N connections by hand. You were cheaper than building the bridge, and more flexible — so the bridge never got built.</p><p>That arrangement came at a real cost. It was capped by how much you could hold at once, so the analysis was only ever as deep as the effort you could put in. The numbers you wrote into a PRD were true the day you wrote them and quietly wrong some days later. And it meant your highest-judgment people were spending hours on clerical reconciliation work that no one would ever thank them for.</p><p>MCP is the thing that completely dissolves this layer. Not by giving you a better text editor or a smarter chatbot, but by letting one AI assistant reach into all those tools — PostHog, Notion, Jira, and the rest — directly, with live access, and do the reconciliation work itself.</p><p>But that raises an obvious question: how does it actually work without handing an AI the keys to everything? That’s where the protocol itself comes in — and it’ll take about 90 seconds to explain.</p><h3>What MCP actually is, in 90 seconds (no USB-C metaphor)</h3><p>I won’t repeat the boring line from all the guides these days: <em>MCP is USB-C for AI.</em> Let’s forget it. It’s a neat slogan that tells you nothing about why your server drops or why your AI slows down — the two things you will actually deal with.</p><p>Here is the version that holds up.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MJZum8Cpq_7zmeMX3anIyw.png" /><figcaption>How MCP works for product managers: the host (Claude, ChatGPT, Cursor) runs the AI and an MCP client that calls MCP servers for Jira, Notion and PostHog, each exposing a menu of allowed read and write actions.</figcaption></figure><p>MCP is the standard way for an AI assistant to discover and use tools it didn’t ship with. That is the whole idea around it. Before <a href="http://anthropic.com/news/model-context-protocol">MCP</a>, whenever you wanted your AI to read Jira, you might need someone to build a custom bridge for Jira specifically — then another for Notion, and another for PostHog. M×N all over again, just moved onto an engineering backlog.</p><p>One simple goal of MCP is to replace all these one-off bridges with one shared protocol, so any tool that “speaks MCP” can be easily picked up by any AI assistant that talks MCP, with no custom wiring between them.</p><p><strong>Three pieces, and you only need to be able to whiteboard these:</strong></p><ul><li><strong>The host</strong> is the app you actually talk to — ChatGPT, Claude, Cursor, or your company’s internal assistant. It is the place where you type and where the AI lives.</li><li><strong>An MCP server</strong> is a connector that sits in front of one of your tools — a Jira server, a Notion server, a PostHog server. It doesn’t expose the raw tool; instead, it presents a short, fixed menu of actions the AI assistant can take, such as “create page” or “search issues”.</li><li><strong>The client</strong> sits inside the host, finds those servers, and brokers the conversation between the AI and the tool.</li></ul><p>So when you ask your assistant to fetch the open bugs from Jira, it does not get handed your Jira password. It asks the Jira server, <em>what can I do here?</em>, gets back the fixed menu of actions, and calls the one that fits. The tool stays in control of what’s on the menu. The AI just orders off it.</p><p>That last point matters most to the users, and it is worth slowing down on. The AI never touches your tool directly. It can only do what each server’s menu allows. That is not a footnote — it’s the entire safety model, and it’s also the reason the gateway problem later in this series exists.</p><p>Two things to keep stored somewhere, that the explainers gloss over. First, the menu for each MCP server can be read-only, or it can let the AI write — in Jira, for example, it can create a ticket, update a record, publish a page. Those are very different levels of risk. And second, every item on the menu costs something to keep running, which is the thread that leads straight to why more tools makes your AI dumber.</p><p>That’s MCP. Not a metaphor — a menu, a waiter, and a kitchen that decides what is cookable.</p><p>But enough about how it works. Let’s go through an example of what it looks like on a normal, messy Monday.</p><h3>Research in one tool, PRD in another — the same Monday, done differently</h3><p>Let’s go back to that PRD you were writing. Same messy Monday, same dozen tabs. Only this time you decide not to open any of them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kmXZJZA3X4MlYUhI3LzIfA.png" /><figcaption>MCP product manager workflow: one prompt to one AI assistant runs research in PostHog, Notion and Jira, drafts a PRD in Notion, then creates Jira tickets — without copy-pasting between tools.</figcaption></figure><p>Instead, you open an assistant and you type something like this:</p><blockquote>“Pull last month’s retention curve from PostHog, find the user interviews in Notion that mention churn, and check Jira for any open bugs tagged onboarding. Then tell me what’s actually going on.”</blockquote><p>And it just… goes. It asks PostHog for the numbers. It runs a search in Notion and reads the interviews. It pulls the open bugs from Jira. And then, at the end, it does the reconciliation work — the part your team used to spend so much time on — except now it happens in minutes.</p><p>This is the key shift. The research that used to eat your morning and a dozen tab-switches now happens in one request.</p><p>Then comes the second one. You say: <em>“Draft a PRD for fixing onboarding churn. Use the numbers you just pulled, and write it into a new Notion page.”</em></p><p>Here is the part that is easy to miss. The PRD it works on is not built on numbers you half-remembered from some random day. It is built on real-time analysis, pulled live while it was writing. Tool A was your research. Tool B is your PRD. And for the very first time, the bridge between them is not you.</p><p>Then the third. You say: <em>“Now break that PRD into Jira tickets — one epic, with the stories underneath it.”</em> It reads the spec it just wrote and creates them. Research in one tool, the document in another, execution in a third — and you didn’t copy-paste a single thing between them.</p><p>Phew. Read that again. The thing that used to eat a third of your week just happened while you watched.</p><p>But I want to be honest about what actually changed here, because it isn’t “the PM got replaced.” Look at what you did in that run. You decided which question was worth asking. You decided this churn deserved a PRD and not just a line in the backlog. You’ll decide which of those tickets ships first. The assistant did the ferrying. You did the judging.</p><p>That is the half of the job worth keeping. The copy-pasting was never the job — it was the tax. MCP just stopped collecting it.</p><p>So that’s the good news, and it’s genuinely good. One assistant, your whole stack, live.</p><p>But nobody who only shows you the shiny demo will tell you the other half. In real life, those servers drop in the middle of a run. Stack enough of them and your assistant gets slower and dumber, not smarter. And the thing that fixes both is a piece of infrastructure most PMs have never heard of — the one I used to build.</p><p>That’s Part 2. I’ll see you there.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0a29445370c2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Agents in Production: The Failure Modes Nobody Puts in the Demo]]></title>
            <link>https://medium.com/@adityadhir97/ai-agents-in-production-the-failure-modes-nobody-puts-in-the-demo-d0b90a4a3ebb?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/d0b90a4a3ebb</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Wed, 03 Jun 2026 15:36:16 GMT</pubDate>
            <atom:updated>2026-06-03T16:09:37.118Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0xjH7aIE1y36YUbnnrN1fQ.jpeg" /></figure><p><em>What separates an AI agent from a chatbot, where they actually pay off, and the security risks — prompt injection, excessive agency, and data leakage — that only surface once you ship.</em></p><p>The only difference between a chatbot and an AI agent is one word, and that word is <strong>act</strong>. An SLM or LLM generates — it takes text and returns text as output. An agent is something that works on that output, whether it’s a tool call, a database query, or sending an email to someone. The model is the mastermind, but now it has hands once you combine it with tools. And the moment you give any software hands, everything changes — how you design and build it, how you reason about it, and how secure it is.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q5ToCMhnAYfjw-kJkOSJUA.png" /></figure><p>I’ve now spent almost a month shipping agentic systems, and most of what people believe about them is hype — and most of the fears are pointed at the wrong things. AI agents aren’t magic, and they’re not going to make you jobless on some random morning. They’re a new generation of software with a new property — autonomy — and that property is the source of both trust and liability. You can’t overweight one over the other. So in this guide, we’ll talk about those tradeoffs.</p><h3>What is an AI agent (and how is it different from a chatbot)?</h3><p>Well, by now you must have come across an AI agent. Still, let’s recap. It’s a new-gen tool that perceives some context, decides what to do next, takes an action, and observes the result — and it repeats this until the goal is met or it just decides to give up. The intelligence isn’t in any single step; it’s in the loop’s ability to adapt when the world doesn’t match the plan.</p><p>And this is what sets an AI agent apart from a chatbot. A traditional chatbot — the kind we’ve all built or used at least once in our lives — follows a set of rules a developer defined in advance, like recognizing a keyword or returning a scripted response. An AI agent is given a goal, and from there it has to figure out a path itself, choosing which tools to invoke and in what order based on what the LLM decides. Anthropic draws the same line in its widely cited <a href="https://www.anthropic.com/research/building-effective-agents">guide to building effective agents</a>: workflows run LLMs and tools along predefined code paths, while agents let the model dynamically direct its own process and tool use.</p><p>That flexibility lets agents handle messy, open-ended tasks that would otherwise take a brittle pile of conditionals to encode by hand. And that’s also why it can be hard to trust them. A rules engine is predictable but not creative; an AI agent is creative but not predictable.</p><h3>How AI agents work: the perceive–decide–act loop</h3><p>If there’s one model to remember behind agents, it’s this: perceive → decide → act → observe, on repeat.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_MAGjuJMsjoGd97aNwCUPw.png" /></figure><ul><li><strong>Perceive:</strong> the agent takes in context — the user’s goal, the current state, and results from previous actions.</li><li><strong>Decide:</strong> the model reasons about what to do next, and which tool to use, if any.</li><li><strong>Act:</strong> it executes an action — a web search, an API call, a database read or write.</li><li><strong>Observe:</strong> it reads the result and feeds it into the next iteration.</li></ul><p>One critical detail a lot of introductions skip: at each step it plans to take, the agent needs ground truth from the environment — actual tool results, real error messages — to know whether it’s making progress. An agent reasoning in a vacuum drifts. An agent that checks its work against reality at each turn is the one you can ship.</p><h3>AI agent use cases that actually pay off</h3><p>The ideal use case for an agent is one that’s repetitive enough to be worth it, but variable enough that it can’t be directly scripted. Some of the patterns that are likely to deliver high ROI:</p><ul><li><strong>Operational automation:</strong> scheduling, triage, data entry, routing requests to the correct application. Agents are good at absorbing the judgment-light coordination work that usually takes up a significant amount of human time.</li><li><strong>Continuous monitoring and synthesis:</strong> watching a stream of information — markets, research, support tickets, logs — and taking action against a standing objective.</li><li><strong>Outbound and follow-through workflows:</strong> good examples are lead qualification, personalized follow-ups, and nudges. Anything where the value is consistent execution but human intervention takes a lot of effort and time.</li><li><strong>Support deflection:</strong> resolving common, well-understood requests end to end, so people only work on the hard ones. This is where teams tend to see their first measurable win, because handling time is easy to track.</li><li><strong>Developer workflows:</strong> code reviews, dependency bumps, repository hygiene, drafting fixes. A lot of engineers are now using AI agents firsthand.</li></ul><p>The common thread across all these use cases: agentic AI is a strong solution where the cost of human attention is high and the cost of a mistake is low. And this brings us to the heart of the article — the part no one puts in the demo.</p><h3>AI agent security risks: the new attack surface</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nVk-Km4x9FUftMHsh3MKTw.png" /></figure><p>The best way I’d put it: unlike traditional software, you’re not securing an application — you’re securing a decision-maker with credentials. Every traditional software risk still applies to the agent, plus autonomy, which is a new layer stacked on top.</p><p>And I’m not speculating. If you check the <a href="https://owasp.org/www-project-top-10-for-large-language-model-applications/">OWASP Top 10 for LLM Applications</a> — one of the most referenced security frameworks for these systems — it now also covers the agentic risks discussed here. These are the failure modes that have made a lot of teams shut down their systems.</p><ol><li><strong>Prompt injection:</strong> a signature agent vulnerability, and OWASP’s number-one LLM risk for the second edition running. Imagine your agent reading a webpage, processing an email, or ingesting a document a user is trying to upload — someone can just add some hidden instructions, which the agent will easily obey, because at the model level it can’t reliably tell data from commands. A support agent summarizing a customer message blindly follows an instruction injected into that message, leaking its system context along with it. This is the well-documented class of <a href="https://arxiv.org/abs/2302.12173">indirect prompt injection</a>, and there’s no clean fix for it — only defense in depth.</li><li><strong>Excessive agency:</strong> if you give an agent broad access to tools to perform more than one task, it can backfire. Some common instances: deleting a record, sending a payment to the wrong person, or publishing something it shouldn’t have. Autonomy plus a whole set of permissions is the ideal combination for turning a small reasoning error into an incident.</li><li><strong>Privilege escalation through tools:</strong> agents usually act through tools, and tools carry a set of permissions. A weakly scoped integration can route the agent to a place it was never supposed to go — a good example is a CRM connector exposing admin-level operations.</li><li><strong>Data leakage:</strong> agents can sometimes see something they shouldn’t. Imagine you’ve asked the agent to write an email to your friend — but what if it sends the email to the wrong person? This is especially dangerous in multi-tenant systems: one tenant’s context must never bleed into another’s.</li><li><strong>Supply chain exposure:</strong> every external tool, plugin, model endpoint, vector database, or third-party API the agent depends on plays a big role in widening the blast radius. A compromised dependency can not only leak data but also steer the agent’s behavior.</li><li><strong>Data poisoning:</strong> imagine giving an agent corrupt information — it just learns from it and uses it at runtime, leading to corrupted decisions. Especially in retrieval-augmented systems, an attacker can simply poison the knowledge base with malicious data, so that when someone asks a question, the agent serves that data right back.</li><li><strong>Weak auditability:</strong> when something goes wrong — and in many cases it will — will you be able to reconstruct what the agent saw, decided, or did? If the answer is no, you might not have a security problem yet, but you also have no way to find one.</li></ol><h3>How to secure AI agents: designing for safe autonomy</h3><p>The good news is that the mitigations are mostly disciplined engineering, not some out-of-the-box research. Teams that ship agents responsibly can easily converge on the same handful of practices — which end up lining up closely with OWASP’s own guidance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DktLUaR8ey-MaTsBjj1Fdg.png" /></figure><ul><li><strong>Least privilege, enforced at the tool boundary:</strong> always make sure the agent has access to the exact tools and scopes it requires to execute a task, nothing more. Scope it per tenant, per user, or per task — whatever suits your product. The tightest agent systems treat every tool call as a permission decision, not a convenience. If the agent doesn’t have a capability in the first place, no prompt injection can succeed.</li><li><strong>Human-in-the-loop gates on consequential actions:</strong> reading is cheap nowadays; acting is expensive. Let the agent reason freely, but put irreversible, high-stakes operations — payments, deletions, external publishing — behind an approval step. The art is to gate the few actions that matter without forcing humans to sign off on thousands that don’t.</li><li><strong>Validation and filtering on untrusted input:</strong> always treat everything the agent ingests from the external world as hostile until proven otherwise. Try to segregate and mark untrusted content so it can’t masquerade as instructions. It’s not always possible to catch every injection, but you can always dramatically shrink the surface.</li><li><strong>Observability as a first-class feature, not an afterthought:</strong> as a matter of good engineering practice, trace every step — the context the agent saw, the decision it made, the tool it used, and the result it got. This turns the tables: from being puzzled about why the agent went weird, to having a debuggable event.</li><li><strong>Dependency hygiene:</strong> always audit the tools and integrations the agent relies on before you productionize it. The agent’s trust boundary should be discoverable, because it’s the union of everything it can call.</li></ul><p>None of these practices aims to make the agent less capable. They make its capability bounded and observable, so that whenever it reasons incorrectly, the failure is contained and you can see it happen.</p><h3>Build your own vs. a managed AI agent runtime</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WXWgbF9WK6xI_oGVxkMkyg.png" /></figure><p>Once you commit to shipping an agent, you hit an architectural fork: assemble the loop yourself, or build on a managed agent runtime.</p><p>Rolling your own gives you total control over the orchestration logic, the execution model, and how state and tools are wired up. Reach for it when the agent’s behavior is your core differentiator and you want to own every layer.</p><p>A managed runtime, on the other hand, can do a lot of the heavy lifting. The good ones handle the full tool-calling lifecycle — invoking the model, calling the tool, feeding the result back into the loop — so you don’t have to write any of it yourself. They manage conversation and execution state, ship a catalog of built-in capabilities like code execution, search, retrieval, and integrations and, critically, give you observability and tracing out of the box. A lot of teams go with a runtime for that last point alone: to get a complete audit trail they’d otherwise have to build.</p><p>It’s also worth knowing the shapes agents come in, because they map to different control needs. Declarative agents are defined by configuration — a model, instructions, tools, and prompts wired together; the simplest are single prompt-based agents, the more ambitious are multi-agent workflows where specialized agents collaborate under an orchestration you describe. Hosted agents go the other way: you write the logic in code and let a platform run it, keeping full control over behavior while offloading infrastructure. Anthropic’s building-effective-agents guide makes the case I agree with: start with the simplest pattern that works, and add agentic complexity only when it earns its keep.</p><p>There’s no universally right answer. The honest version of the decision: how much of this agent’s behavior is your moat, and how much is plumbing you’d rather not maintain?</p><h3>The one-sentence version</h3><p>If you take a single idea from this, take the verb. Agents act. That’s the entire story — the reason they’re powerful, the reason they’re risky, and the reason you should treat them like any system you’d let make decisions on your behalf: scope its power, watch what it does, and gate the actions you can’t take back.</p><p>The teams that internalize this aren’t the ones moving slowly. They’re the ones still standing after their agents meet the real world.</p><h3>About the Author</h3><p>Aditya Dhir is a Senior Backend and Deep Technology Engineer focused on building production-grade AI systems, agentic workflows, cloud infrastructure, and distributed architectures. He writes about AI engineering, system design, multi-tenant SaaS platforms, and lessons learned from building real-world software.</p><p>Connect with me on LinkedIn:</p><p><strong>LinkedIn:</strong> <a href="https://linkedin.com/in/adityadhir">https://linkedin.com/in/adityadhir</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d0b90a4a3ebb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Almost Shipped an AI Agent That Could Have Exposed Customer Data]]></title>
            <link>https://medium.com/hackernoon/i-almost-shipped-an-ai-agent-that-could-have-exposed-customer-data-af1c5a750efd?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/af1c5a750efd</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Fri, 29 May 2026 16:07:35 GMT</pubDate>
            <atom:updated>2026-05-29T16:10:03.077Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Multi-tenant AI agent architecture where one shared agent instance serves two tenants and leaks cross-tenant customer data." src="https://cdn-images-1.medium.com/max/1024/1*Fu6MIdAL8mOxn1sIPWRVhg.png" /><figcaption>Multi-tenant AI agent cross-tenant data leak — one shared agent, two tenants</figcaption></figure><h3>The hidden state-management problem lurking inside multi-tenant AI systems.</h3><p>Everything was working perfectly. Unit tests were passing, smoke tests were clean, and the dev environment was rock solid. Then, during a final round of concurrency testing, I fired two requests simultaneously — and saw something that stopped me cold.</p><p>Two different tenants. The exact same data.</p><p>This wasn’t something discoverable through standard unit tests or casual local development. We only caught it because I happened to script two separate tenants hitting the exact same workflow at the exact same millisecond. That’s the moment I pulled the deployment.</p><h3>What We Were Building</h3><p>The product was an AI-native layer built on top of an IT Service Management (ITSM) platform. The premise was simple: a user types a request in plain English — for example, <em>“The VPN keeps dropping for the finance team; can you log this?”</em> — and the agent automatically transforms it into a properly structured support ticket. It assigns the right category, sets the correct priority, and handles routing using context pulled from the tenant’s specific knowledge base.</p><p>Like most B2B SaaS applications, this was a multi-tenant system. One deployment, many customers, each completely walled off from the next. In our world, tenant isolation isn’t just a checklist feature — it is the bedrock foundation of the product. Under no circumstances should Tenant A ever catch a glimpse of Tenant B.</p><p>That is the line we came within one deployment of crossing.</p><h3>Why Everything Looked So Fine</h3><p>The most terrifying part of this bug was that our standard testing pipeline was completely blind to it.</p><ul><li><strong>Unit tests</strong> run one request at a time. One input, one execution path, one isolated assertion. They are fantastic at proving a function works correctly when it’s the only thing running on the CPU. They cannot tell you what happens when two requests are in flight at the same moment, quietly stepping on each other’s toes.</li><li><strong>Smoke tests</strong> suffered from the exact same blind spot. They confirmed that the happy path worked, not that two happy paths could safely coexist simultaneously.</li><li><strong>Local dev</strong> “felt” incredibly stable because it was just me, sitting in a silo, clicking through features one by one.</li></ul><p>We had built a system that was perfectly correct in isolation, but profoundly dangerous under load. And every single test we wrote was designed to look at it in isolation.</p><h3>The Moment It Broke</h3><figure><img alt="Race condition diagram showing a shared AI agent’s tenant_id field overwritten by a concurrent request, causing a cross-tenant data leak" src="https://cdn-images-1.medium.com/max/1024/1*vta8-jC0K0Oggmq0wDqGuA.png" /><figcaption><em>AI agent race condition — shared tenant state overwritten under concurrency</em></figcaption></figure><p>The breakthrough came when I stopped testing the agent like a single, well-behaved user and started testing it like a high-traffic platform. I wrote a script to simulate two completely separate tenants hitting the exact same workflow at the exact same time: Tenant A and Tenant B, firing interleaved concurrent requests.</p><p><strong>Tenant A’s response came back populated with Tenant B’s data.</strong></p><p>This wasn’t intermittent flakiness I could shrug off or blame on network jitter. It was a clean, reproducible, cross-tenant leak that only materialized when two requests overlapped in time. Which is to say, it was exactly the condition that would exist every second of every day the moment we launched in production.</p><h3>The Root Cause: Shared State in a Hidden Corner</h3><p>So, how did we build a cross-tenant data leak? It came down to a classic optimization choice: reusing a single agent instance.</p><pre>[Request A] ───►  Assign Tenant A  ──┐<br>                                     ▼<br>                               [Shared Agent]  ──► [Reads Tenant B Context!] ──► Leak!<br>                                     ▲<br>[Request B] ───►  Assign Tenant B  ──┘</pre><p>Agents are computationally expensive to spin up. They have to pull down system prompts, initialize tools, load embedding models, and assemble foundational context. To avoid paying that latency penalty on every API call, we instantiated the agent once at startup and reused it to handle incoming traffic. It felt like the obvious, efficient thing to do.</p><p>The foundational flaw was that our agent instance was stateful — it held the active tenant’s context directly on itself.</p><p>Under sequential traffic, this design flaw masqueraded as a feature. Request A would finish, and then Request B would come in and overwrite the context. No harm done. The bug never had the breathing room to exist.</p><p>But concurrent traffic turned that assumption into a time bomb. Request B set its tenant context on the shared agent while Request A was still mid-flight. When Request A went to read its context back to finalize the ticket, it grabbed Tenant B’s data. Whoever wrote last won, and the other tenant got breached.</p><h3>Why Multi-Tenant AI is Uniquely Vulnerable</h3><p>In traditional CRUD applications, the stateful surface area is small and deeply audited. Developers learn early on to explicitly scope database queries by a tenant ID.</p><p>But AI agents are complex machines. They naturally accumulate state across multiple layers:</p><ul><li>System prompts and dynamic prompt assembly</li><li>Vector database retrieval layers</li><li>Conversation memory stores</li><li>Tool-calling glue and scratchpads</li></ul><p>Every single one of these layers is a slot where tenant-exclusive data lives. When any of these components are tied to a shared, long-lived object instead of being scoped strictly to a single execution thread, you don’t just get a minor software bug. You get a natural language data breach.</p><p>Worse yet, because the output is fluid, polished prose rather than a raw database error, a leak doesn’t look like a failure. It looks like a perfectly normal, well-written response — it just happens to belong to the wrong company. The system won’t throw a 500 error; it will just smile and hand away your data.</p><h3>The Fix: Moving Tenancy to the Tool Boundary</h3><figure><img alt="MCP tenant-scoped tool calling diagram with a stateless AI agent and per-request tenant isolation enforced at the tool boundary" src="https://cdn-images-1.medium.com/max/1024/1*9IyiNiZUFTAfMzuMPBYXUw.png" /><figcaption>Tenant scope rides with every call: the agent holds no tenant state, and the MCP boundary enforces isolation per request.</figcaption></figure><p>The immediate, emergency fix was straightforward: stop sharing the agent. Giving every single request its own short-lived instance eliminates the shared slot entirely. If you need to patch a leak tonight, do that.</p><p>But treating that as the permanent solution is just a band-aid. It fixes <em>this</em> specific leak by isolating the object, but it doesn’t prevent the next engineer from accidentally parking tenant data on a different shared global utility down the road. We wanted tenant isolation to be a foundational property of our architecture, not a rule everyone on the team had to remember to enforce perfectly every time.</p><p>To solve this permanently, we moved tenancy off the agent entirely and pushed it down to the tool boundary using <strong>Model Context Protocol (MCP)</strong> and tenant-scoped tool calls.</p><p>Now, the agent is completely stateless regarding tenancy. It reasons, it plans, and it decides which tools to call, but it holds zero proprietary tenant data in its own memory. Instead, every single tool request it initiates carries the tenant scope with it dynamically, and the MCP server enforces that scope on a per-call basis.</p><pre>                     ┌───────────────────┐<br>                     │   Stateless LLM   │<br>                     │    Agent Loop     │<br>                     └─────────┬─────────┘<br>                               │<br>               Executes Tool Call + Tenant Token<br>                               │<br>                               ▼<br>              ┌─────────────────────────────────┐<br>              │  MCP Server / Tool Boundary    │<br>              │  (Enforces Tenant Verification) │<br>              └────────────────┬────────────────┘<br>                               │<br>                        Isolated Query<br>                               │<br>                               ▼<br>                     ┌───────────────────┐<br>                     │  Tenant Database  │<br>                     └───────────────────┘</pre><p>Scope rides with the request, not the object. Two concurrent calls for two different tenants are processed as two independently scoped payloads. They can never overwrite each other because they never share a common memory slot.</p><p>By enforcing this at the boundary where data is actually touched, we stop trusting the LLM to “stay in its lane.” Even if the agent’s reasoning loop encounters a bug, or the model hallucinates a request for the wrong account, the underlying MCP tool physically cannot access or return another tenant’s data.</p><h3>Our Multi-Layered Defense</h3><p>Because the core lesson here is that any single layer can fail silently, we backed this up with two additional guardrails:</p><ul><li><strong>Data-Layer Isolation:</strong> We enforce tenant IDs on every database query alongside row-level security (RLS), meaning the database itself will explicitly refuse a cross-tenant read.</li><li><strong>A Last-Line Assertion:</strong> Before any payload is dispatched back to the user, a final middleware interceptor verifies that every record’s tenant ID matches the requesting user’s tenant ID. If there’s a mismatch, the system fails loudly and aborts the response entirely.</li></ul><h3>The Ultimate Lesson</h3><p>The fix required only a few lines of code, but the architectural lesson is massive: <strong>Green tests do not mean safe code; they simply mean correct-in-isolation.</strong> For a multi-tenant system under real-world conditions, those are two entirely different claims.</p><p>Concurrency and tenant-isolation testing are no longer afterthoughts run on a rogue hunch — they are first-class citizens in our CI/CD pipeline. Every single change we commit automatically fires multiple simulated tenants at the same endpoints at the exact same time to assert zero cross-contamination. We test the system like a platform under load, because production is never a single, well-behaved user.</p><p>If you are currently building or deploying multi-tenant AI agents, do this before you ship: write a script, throw two tenants through your most stateful workflow simultaneously, and check — <em>really check</em> — whose data comes back.</p><p>I am incredibly glad I did. The alternative was finding out via a customer support email.</p><h3>A quick question on your implementation:</h3><p>When you shifted to the stateless agent model with MCP, did you experience any noticeable latency hit from passing the tenant scope and spinning up tenant-scoped credentials on every tool call, or did the performance remain relatively stable compared to the shared-instance approach?</p><h3>About the Author</h3><p>Aditya Dhir is a Senior Backend and Blockchain Engineer focused on building production-grade AI systems, agentic workflows, cloud infrastructure, and distributed architectures. He writes about AI engineering, system design, multi-tenant SaaS platforms, and lessons learned from building real-world software.</p><p>Connect with me on LinkedIn:</p><p><strong>LinkedIn:</strong> <a href="https://linkedin.com/in/adityadhir">https://linkedin.com/in/adityadhir</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=af1c5a750efd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/hackernoon/i-almost-shipped-an-ai-agent-that-could-have-exposed-customer-data-af1c5a750efd">I Almost Shipped an AI Agent That Could Have Exposed Customer Data</a> was originally published in <a href="https://medium.com/hackernoon">HackerNoon.com</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[5 Hard Lessons I Learned Building an AI-powered Backend (That Nobody Warned Me About)]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/hackernoon/5-hard-lessons-i-learned-building-an-ai-powered-backend-that-nobody-warned-me-about-810fcdcaf797?source=rss-34c84bf846ce------2"><img src="https://cdn-images-1.medium.com/max/1732/1*nxfpdIy8wu-OY_yXRcuUfQ.png" width="1732"></a></p><p class="medium-feed-snippet">What months of shipping a multi-tenant AI-enabled SaaS actually taught me was just beyond the tutorials I found.</p><p class="medium-feed-link"><a href="https://medium.com/hackernoon/5-hard-lessons-i-learned-building-an-ai-powered-backend-that-nobody-warned-me-about-810fcdcaf797?source=rss-34c84bf846ce------2">Continue reading on HackerNoon.com »</a></p></div>]]></description>
            <link>https://medium.com/hackernoon/5-hard-lessons-i-learned-building-an-ai-powered-backend-that-nobody-warned-me-about-810fcdcaf797?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/810fcdcaf797</guid>
            <category><![CDATA[ai-engineering]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Tue, 26 May 2026 12:03:06 GMT</pubDate>
            <atom:updated>2026-05-26T12:03:06.296Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[What all you need to know about Amazon web services or AWS?]]></title>
            <link>https://medium.com/@adityadhir97/what-all-you-need-to-know-about-amazon-web-services-or-aws-38bf49689a1?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/38bf49689a1</guid>
            <category><![CDATA[amazon]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Thu, 30 Jul 2020 11:17:54 GMT</pubDate>
            <atom:updated>2020-07-30T11:17:54.289Z</atom:updated>
            <content:encoded><![CDATA[<h3>Introduction.</h3><p>With technology advancement, cloud computing has become one of the emerging technologies. Due to pandemic, many people are shifting towards work from home and they do find the need for some cloud computing services to perform some computations and share data with others while managing a project.</p><p>With such a huge demand for cloud computing services, different companies have come up with their cloud computing platforms. Some of the famous cloud computing platforms are Google Cloud, AWS and Microsoft Azure, and many more.</p><p>AWS or Amazon web services is a cloud computing platform by Amazon.com. Amazon came up with the idea of a cloud computing platform when they decided to rent out extra servers that they bought for storing amazon’s e-commerce website data. Such services are really useful when you need to reduce costs and build anything without hassle. Such cloud computing software is very useful for users.</p><h3>What is Amazon Web Services?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2368t-zTSVzwfBQQ" /></figure><p>According to the official website of Amazon, it is a cloud computing platform for users and business owners to conduct different powerful computations remotely on Amazon servers. Amazon web services or AWS is the world’s largest cloud computing platform with around 175 full features services. Several data centers are located globally. Around millions of customers including big businesses, startups, and government agencies use AWS. it is proven to be lower in cost, more agile, and innovative.</p><h3>Different Functions of AWS.</h3><p>AWS has a huge number of services along with different features. Such a platform lets users compute, store, and work on emerging technologies like Machine learning and Algorithm. It does provide several applications to run data analytics and the Internet of things. It is considered to be more fast and cost-effective.</p><p>Since cloud computing does use the internet, hence security is a major concern. AWS is designed in such a way that it is more flexible and secure for cloud computing. All such applications available in AWS are do fulfill the requirements of high sensitive companies, global banks, and the military. It does have the backing of a large number of security tools with a huge number of features and governance services. AWS does follow 90 Security standards along with different compliance certifications. It stores all the data in its 117 AWS services with encryption.</p><p>Such a cloud platform has shown to be really innovative. It lets users to several latest technologies and build innovative products. Let us take an example, in 2014 AWS developed a serverless computing space and launched AWS lambda. Such a feature allows users to run their code with server management. Similarly, Amazon Sagemaker is a fully managed machine learning service allowing developers to use machine learning.</p><h3>Different solutions By Amazon AWS.</h3><p>Amazon AWS provides a wide range of solutions on its cloud computing platform. Such solutions include various new technologies.</p><h3>Machine Learning.</h3><p>AWS is built with a wide range of powerful services and different platforms. Such a solution helps users to implement the broadest machine learning framework from anywhere. You can build different ML-based codes on AWS and do the trained model on AWS. it would reduce the overall cost.</p><h3>Analytics and Data Lakes.</h3><p>AWS allows users to securely store, analyze, and categorize all the data in a centralized repository. As you know data volumes are increasing day by day and the huge cost is involved in storing and analyzing such amount of data. AWS provides a best and cost-effective tool that lets users store and analyze all the data. Different database solutions are available in AWS for data.</p><h3>Internet of things.</h3><p>Several devices are available in homes, factories, cars, and many more. With increase in the number of devices, solutions need to connect them increases. Several solutions let users collect analyze and store such data. AWS provides users all the security layers and different security mechanisms like encryption and various controls.</p><h3>Serverless computing.</h3><p>Serverless computing is a new technology that enables users to build different modern applications with more agility and low cost. Such serverless applications would let users to more focus on the core product instead of managing and building different servers. Such would reduce the cost and time. Developers would be able to focus more on the product and build a reliable product.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*SIyXXJLIRGBQyQlj" /></figure><h3>Why should you use AWS services?</h3><p>Such services are designed in the best way and built on four main factors.</p><h3>Security.</h3><p>AWS is built on different comprehensive security capabilities that satisfy different users and various requirements. Different security measures are taken by AWS.</p><h3>Compliance Free.</h3><p>It gives users different features such as rich controls, broad security accreditations, and auditing.</p><h3>Hybrid.</h3><p>It lets users build different hybrid architectures that let users extend different on-premises infrastructure on Cloud.</p><h3>Scalable.</h3><p>Amazon AWS lets users build scalable applications. Such applications can be scaled up and down according to requirements.</p><p>Amazon AWS services have let different users build some great applications. Along with it, several other services are available on Amazon for e-learning.</p><h3>DIfferent Amazon E-Learning Services.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/812/0*gq2cQppCFysulssA" /></figure><h3>Amazon Events and Webinars.</h3><p>Amazon hosts different kinds of events and webinars for everyone to learn more about cloud computing and build a community that can connect, learns, and collaborate with different AWS experts. Such events are held for both experienced and newbies, it let uses to learn more about AWS and build some great skills.</p><h3>AWS training and certifications.</h3><p>Different organizations do need various skilled individuals to transform business. Different AWS certifications and training lets anyone build and validate different cloud skills. All the content for such training and certifications are according to all different services by AWS.</p><p>AWS holds both digital and classroom training. Such training allows users to learn remotely as well as in the classrooms.</p><h3>AWS partner network.</h3><p>AWS partner network is a global program for partners who have a great interest in technology and consulting and wants to leverage Amazon web services for building solutions for customers. Such network assists different companies to build, sell, and market different AWS offers with valuable marketing, technical and business support.</p><p>Different partners are available across the globe and many companies are fortune 100 and fortune 500. All such businesses using Amazon.</p><h3>Building applications with Amazon web services.</h3><p>Amazon web services allow users to deploy their applications on services. If you are looking to deploy any such application on the Amazon Web Services platform, you need to make sure that few things are taken care of. It is really important to know that deploying applications in AWS do not require an underlying architecture.</p><h3>Different benefits of amazon web services and cloud computing.</h3><ul><li><strong>Amazon web services reduce cost.</strong> Since business now does not need to spend huge finances on data servers and data centers. You need to pay only for the resources which you use.</li><li><strong>It brings a more scalable economy.</strong> With the use of amazon web services, you can easily achieve a lower cost and cover it on your own. Many users are working on cloud and AWS lets users achieve high scalability in the economy. Hence it would require fewer finances.</li><li><strong>You don’t need to worry about infrastructure storage</strong>. As you are allowed to use, how much you require to deploy any application. Sometimes it does cost high if resources are selected wisely. Cloud computing does not face any such issues in storage. Users can easily scale it according to use.</li><li><strong>Amazon web services increase the speed and agility of the application.</strong> It means all the IT resources are just one click away. You can easily increase the agility of the software and cost is reduced.</li><li><strong>Amazon web services do not let users spend a huge amount of money on infrastructure and maintenance of different data centers</strong>. It let AWS users to focus more on business and its product rather than infrastructure. AWS would let more focus on business rather than heavy servers.</li><li><strong>With Amazon web services you can easily deploy your application in several rights around the world with just few clicks.</strong> It would reduce the latency and give customers access at a minimal cost.</li></ul><h3>More About AWS</h3><p>Amazon web services are a great platform in the field of cloud computing and several users do use it for businesses. Amazon would let you work on different facilities and services at a very low cost. If you are a business owner and would like to use it for your business, you should go for amazon web services.</p><p>If you are a learner and would like to learn about amazon web services, you can access different courses available by several websites and Amazon online. Several tutorials would assist to you know amazon web services and you can use it for your business, educational work, and much more. For researchers, it would be a great choice since it will be useful for learning about the cloud and how does it work. Researchers can work on different research work based on new technologies and do not require any high configuration devices.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=38bf49689a1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Latest technologies in 2020: What all you need?]]></title>
            <link>https://medium.com/@adityadhir97/latest-technologies-in-2020-what-all-you-need-c76503b4f1f9?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/c76503b4f1f9</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Mon, 13 Jul 2020 05:19:46 GMT</pubDate>
            <atom:updated>2020-07-13T05:19:46.029Z</atom:updated>
            <content:encoded><![CDATA[<p>With the rise of global pandemic in 2020, many people have beginned using different technologies for their businesses, education, work and much more. Due to social distancing and prevention of covid19 spread, many people have shifted from cubicles to home office, classrooms to virtual classroom. In a short span of time many people have adopted different use cases of technologies.</p><p>And in order to make work less affected due to global corona virus pandemic. It has all become possible due to emerging technologies, like artificial Intelligence, Blockchain, cyber security, eCommerce. And different tech tools like online meeting applications and websites, collaborative applications of work and much more.</p><p>Read More:<a href="https://techiesphere.co/blockchain/latest-technologies-in-2020-what-all-you-need-to-know/">https://techiesphere.co/blockchain/latest-technologies-in-2020-what-all-you-need-to-know/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c76503b4f1f9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Quantum Computing: the Future of computation!]]></title>
            <link>https://medium.com/@adityadhir97/quantum-computing-the-future-of-computation-acd18d4f9850?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/acd18d4f9850</guid>
            <category><![CDATA[quantum-physics]]></category>
            <category><![CDATA[quantum-computing]]></category>
            <category><![CDATA[quantum-mechanics]]></category>
            <category><![CDATA[quantum-computer]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Sun, 21 Jun 2020 05:41:20 GMT</pubDate>
            <atom:updated>2020-06-21T05:41:20.075Z</atom:updated>
            <content:encoded><![CDATA[<p>For the past few centuries we have seen a huge evolution of computers and we are in an era where we are using computers for all basic purposes. It could be mobile phones, laptops, desktops.but the problem with the computers we use, still takes time for doing huge and complex computation. We need a superfast computer which could do huge computations at a very less time. Imagine what if we can build a computer on the concepts of Quantum mechanics. Well, it would be really fast at doing computations. And these computers would be known as quantum computing machines.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*dIWTGOAKakZbin5b" /></figure><p>Image src<a href="https://atos.net/en/insights-and-innovation/quantum-computing">:https://atos.net/en/insights-and-innovation/quantum-computing</a></p><h3>A basic difference between Quantum computers and basic computers.</h3><ul><li>Well, the systems we use now, information is stored in the form of bits, but in quantum computers, information is stored in the form of qubits also known as quantum bits.</li><li>A Basic Principle behind quantum computers is that for data representation we can use quantum properties. Also, various quantum mechanics would be used to perform various operations in data.</li><li>Imagine what if we are able to build quantum computers, will it break all the cryptography concepts and be a superfast system which will perform computations in less time. We will discuss all the important aspects of Quantum computing in the blog.</li></ul><h3>Quantum computers still at the age of infancy.</h3><p>What all computations are done by quantum computers are performed on very fewer qubits. Various theoretical and practical research is going on at a frantic pace. Various agencies, such as government and military people are working on quantum computing research. Good research is going on for various security purposes such as civilian security and national security.</p><p><strong>Quantum mechanics concepts like “superposition” and Entanglement.</strong></p><p>There is no doubt that quantum computers are using concepts of quantum mechanics such as superposition and entanglement. It is fine if you are baffled with concepts like these. Well, we have various quantum particles like electrons, photons, atoms.</p><ul><li><strong>What is superposition?</strong></li><li>Superposition is the property of a quantum system that lets the system to be in multiple states at the same time. You can think of this like, a system which could be “here” and “there” or “up” and “down”.</li><li><strong>Entanglement.</strong></li></ul><p>Well, entanglement builds a strong correlation between quantum particles. Such as two or more quantum particles could be linked together inextricably even they have huge distance between them.</p><p>All the particles are linked together in such a way that, it seems that they dance even if there is a huge distance between them such as both or many particles are located at different ends of the universe. Also, this observation of long-distance connection inspired einstein to present this concept as “ spooky action at a distance”.</p><h3>Why these quantum are significant?</h3><p>Well, it is really fascinating. And it could be really useful for doing computations and communications. Well, special thanks to the concepts like superposition and entanglement since these concepts help to do huge calculations simultaneously. We all know simple computers work on ones and zeros, but quantum computers work on ones, zeros, and superposition of ones and zeros. Various difficult tasks which are impossible to perform on simple computers could be performed by quantum computers in very little time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PnSdp4Aoa9SkBcz_" /></figure><p>Image source: <a href="https://atos.net/en/insights-and-innovation/quantum-computing">https://atos.net/en/insights-and-innovation/quantum-computing</a></p><h4>What quantum computers are capable of doing?</h4><p>Well, with the advancement in quantum computing. They have started to approach the classical simulations. Various basic operations like factorization of large numbers, and multiplying huge numbers would be easy for quantum computers if digits are of a huge number. Also according to a mathematician named Peter Shor, who was working in AT&amp;T claimed that if a quantum computer would be available, it could easily factor numbers with a huge number of digits.</p><p>Well, no person would spend huge amounts of money to build quantum computers for working on factors of numbers or much more. Factors of numbers are somewhere related to cryptography concepts. Such as in RSA we use factors of the number to encrypt data.</p><p>With a classical system that we have now would not be able to decrypt the information encrypted by RSA. Since it needs a key which is two prime numbers and their multiplication for decryption. No eavesdropper could ever do this if one does not build a quantum computer and implement some of Peter Shor’s algorithm. So, well does it mean Quantum computers could hack all data?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*W4Wq1h4nZFvlEa9L" /></figure><p>Image src: <a href="https://www.thesslstore.com/blog/post-quantum-cryptography-10-things-you-need-to-know/">https://www.thesslstore.com/blog/post-quantum-cryptography-10-things-you-need-to-know/</a></p><ul><li><strong>Various Advantages of Quantum computing.</strong></li><li>So, the use of peter Shor’s algorithm has been a motivation for people to do good research in the field of quantum computers. Well, in Broader view quantum computers could perform various solutions to huge problems in very less time.</li><li>Various researchers are working on several problems that could be solved in a few seconds. Also, researchers are trying to find ways to develop speedy algorithms.</li><li>Various fields with future scope of research are quantum networks and communications, quantum sensing, and metrology. Well, the use of quantum computers would lead to huge revolutionary ways for information sharing.</li></ul><h4>What all the facts you should know about quantum Computing?</h4><p>Like earlier said, Also, quantum computers would change the whole landscape of data security.</p><p>Encryption algorithms we do have know would become easily breakable with quantum computers. It has the potential to crack all encryption techniques and various predictions have been made that they would build hack-proof algorithms.</p><p>Classical computers are better than quantum computers in various tasks like (email, desktop, spreadsheets and much more). Well, the main aim of building quantum computers solves various problems but it won’t be replacing classical problems.</p><p>Quantum computers would be great for solving optimization problems like booking flight tickets or finding the best route for FedEx.</p><p>According to Google, Quantum computers would be a million times better than a classical computer on google.</p><p><strong>Some startups working on quantum computing.</strong></p><ul><li><strong>1Qbit.</strong></li><li>It is a software and consulting company working on quantum computing. And It aims to provide some solutions which could be able to solve various difficult problems using a complex algorithm. It will use various software development tools and various classical methods along with quantum computers.</li><li><strong>A*Quantum</strong></li><li>It is a company build in 2018, a Japanese company that works on quantum computer software. Also, it develops software that would be useful for both gate-level machines.</li><li><strong>SoftwareQ</strong></li><li>It is another company that is working on quantum software and provides a good number of services which include various simulators, optimizers, education software, and consultancy services. It is a Canadian company build in 2017.</li><li><strong>Qulab</strong></li><li>A company founded in 2017 in California.it aims to automate molecular drug design with the use of classical and quantum computers.</li></ul><p><a href="https://quantumcomputingreport.com/players/privatestartup/">For more details on startups.</a></p><h4>IBM and Quantum Computing</h4><p>Well, in the current era, IBM is leading in quantum research. It has got a special division called IBMQ, it is a special company working on quantum computing to make it accessible widely. Also, it could shed a good amount of light on what the future holds for quantum computers. It provides a Qisikit extension. With this, it has got an open-source program that allows several developers and researchers to explore the whole concept of quantum computers with the help of python scripts. Also, it allows people to collaborate on various requests to interactions of quantum computing.IBM aims to make breakthroughs in quantum computing. And the initial initiative was done by Richard Feynman.</p><h4>Final Words</h4><p>Quantum computers have gotten exponentially large advantages over classical computers. Also, It would revolutionize the world, implemented on a large scale.</p><p>Hence,Various Quantum technologies would bring a second quantum revolution. From industries would have a huge impact on Quantum computing.</p><p>It would change aspects of life. Also, it would bring a huge explosion of data generated using IOT and Big data. It would bring advancement in artificial intelligence and deep learning.</p><p>Hence, Quantum computing would bring major change cryptography concepts and it will make various technical updates in the coming decades.</p><p>Well, in a future blog we will discuss deep concepts in Quantum computing.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=acd18d4f9850" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is Education a business nowadays?]]></title>
            <link>https://medium.com/@adityadhir97/is-education-a-business-nowadays-828b15a5291d?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/828b15a5291d</guid>
            <category><![CDATA[education]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[education-technology]]></category>
            <category><![CDATA[industry]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Mon, 15 Jun 2020 14:09:45 GMT</pubDate>
            <atom:updated>2020-06-15T14:09:45.510Z</atom:updated>
            <content:encoded><![CDATA[<h3>Introduction</h3><h4><a href="http://techiesphere.co/">“Education is not a Debt sentence”- Anonymous</a></h4><p>“Education is not a debt sentence” is what is written on the image I recently saw. This line has made me think in a much broader way and forced me to realise that education has become a business and converted into commodity. Well, yes in the name of education, a huge business is running and some of the successful stakeholders are private coaching institutes, schools and colleges. It is too sad to believe that education is just a commodity now and to get that commodity anyone is ready to give all the savings and everything.</p><figure><img alt="Education is a basic right" src="https://cdn-images-1.medium.com/max/819/0*UMqOTgPvM6nlV2XK" /></figure><p>From school to cracking an entrance exam to getting admission in the college, everyone just pays! People who can afford easily pay but what about the people who just can’t afford?</p><h3>History of Education</h3><p>In ancient India, nothing was the same back then, education was the right for everyone and “guru Dakshina” was something that all the shishyas gave to their gurus for teaching them. Everyone used to receive education, despite their social status. No fees were charged and the guru( a teacher) used to take care of everything for students(shishya). Education was not a business back then and several education providers were not money minded and very scarifying,, unlike modern times. And Education was a priceless service that everyone received despite their social status. Education was not commodity or business.</p><h3>Education in Modern days.</h3><p>Gone are those days and the concept of guru Dakshina does not exist. Several Education providers are charging a great amount of money for providing education. Education has become luxury and people who can afford that luxury is able to get educated. And one who cannot afford that luxury stays uneducated. Well, it doesn’t sound good but it is a harsh reality.</p><p>According to several statistics, in India, alone millions of people cannot go for education just because they can’t afford to go to school. With the rise in the huge population, and under capitalism, education has become a business, now it is all about making huge profits.</p><p>In recent few years, huge privatisation of education has taken place, people have simply accepted that. It has affected a big group of people who just can’t afford it. Several deserving candidates just need to drop their education because of financial constraints. People who give great amounts of money are welcome in such institutes. Despite having the right to education for everyone, many of them are not receiving education. And that’s why many citizens do stay unemployed.</p><h3>Everyone should have access to education.</h3><p>With the acceptance and letting institutions charge us huge money has turned up education into a business. And somehow the quality of education is not up to the mark, now the focus is more on gigantic buildings and infrastructure and getting a large number of students. Somehow we are also responsible for this because we believe that the more fees they charge, the better education they provide and facilities. The high-quality infrastructure and world-class facilities attracts most of us, and we get ready to pay fees. And to acquire all such facilities many such people take loans and go under debt. It is a harsh truth and we should realise it soon.</p><h3>Conclusion.</h3><p>Education should not be like a commodity or a business. Education is a basic necessity which everyone should receive despite their social status. And Education should be available at a certain cost that is affordable by all. Many uneducated and poor parents also want to see their children as educated. But due to lack of money, they are not able to send. More government schools and colleges should construct and education quality should be good. So that the private sector is not more under monopoly and not in a position to exploit us.</p><p>For more such articles subscribe to our blog — <a href="https://www.techiesphere.co/education">techiesphere.co</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=828b15a5291d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cryptocurrency exchange: A beginner’s guide.]]></title>
            <link>https://medium.com/@adityadhir97/cryptocurrency-exchange-a-beginners-guide-4095257fa3b?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/4095257fa3b</guid>
            <category><![CDATA[binance]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[cryptocurrency-exchange]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[cryptocurrency-investment]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Sat, 13 Jun 2020 13:28:38 GMT</pubDate>
            <atom:updated>2020-06-13T13:28:38.941Z</atom:updated>
            <content:encoded><![CDATA[<h3>Introduction</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/0*Tszbo8Stt2O-5zYC" /><figcaption>Cryptocurrency exchange: A beginner’s guide.</figcaption></figure><p>With the rise of several <a href="https://techiesphere.co/">cryptocurrencies</a> in the industry, many people have started taking interest in doing transactions, investment, and trading cryptocurrencies. Since the inception of bitcoin, various cryptocurrencies exchanges have built. Such a cryptocurrency exchange allows users to do all the activities related to cryptocurrencies. Cryptocurrency exchanges can be a web application, mobile application, or both. Some of the examples of cryptocurrency exchanges are Binance, Coinbase, and many more.</p><h3>What is a cryptocurrency exchange?</h3><p>Cryptocurrency exchanges are web-based or mobile-based applications that allow users to conduct different activities like sending, receiving and trading of cryptocurrencies. Such platforms may or may not have a native currency that could be used to buy different cryptocurrencies. For example, Finance a cryptocurrency exchange does have a native cryptocurrency, BNB. BNB is used by users as a trading pair with different cryptocurrencies</p><p>Cryptocurrency exchanges can be of different types such as a public cryptocurrencies exchange or private-public cryptocurrencies exchange. Such exchanges would be available for only specific people on an invite based whereas public cryptocurrencies exchanges are available for everyone.</p><h3>Centralized cryptocurrency exchanges and decentralized cryptocurrency exchanges.</h3><p>Cryptocurrency exchanges can be categorized further as centralized and decentralized based on how they are controlled. It means that centralized cryptocurrency exchanges are controlled by a central authority. Whereas decentralized cryptocurrency exchanges are controlled by no central authority. It is more distributed and everyone has equal authority over exchanges.</p><ul><li>Centralized Cryptocurrency Exchanges.</li><li>Centralized cryptocurrency exchanges are online platforms used for transacting cryptocurrencies. All the centralized cryptocurrency exchanges do make use of third parties to control and authorize transactions.</li><li>Such cryptocurrencies exchanges allow trading from fiat to cryptocurrency or vice versa. It covers all the important aspects of the market.</li><li>In centralized cryptocurrency exchanges, users do store all the currency in the exchange’ storage. Hence cryptocurrency exchange has control over all the user’s accounts. You must have heard about different stories about centralized exchange getting hacked.</li></ul><h3>Centralized Cryptocurrencies exchanges have several disadvantages.</h3><p>In recent times, several cryptocurrencies exchanges have hacked including binance. Such cyber-attacks include phishing ransomware and many more. It has led to the loss of a huge amount of cryptocurrencies of users.</p><p>Back in 2019, hackers have hacked binance, and around $40 million bitcoins were stolen from the platform. It is the world’s largest cryptocurrency exchange. Such hacks are more likely to occur in centralized cryptocurrency exchanges.</p><h3>Several governments have banned centralized cryptocurrency exchanges in Countries like China, South Korea, and Russia.</h3><p>Centralized Cryptocurrency exchanges do compromise of privacy as they do need different sensitive information of the user. User needs to provide different information such as bank account details, government-issued IDs, and much more.</p><p>In centralized exchanges, people do not have access to the private keys of exchange account wallets hence, all traders need to trust exchange operators. All the transactions are through the common mechanism by a central authority.</p><ul><li>Decentralized cryptocurrency Exchanges or DEXs.</li><li>Decentralized cryptocurrencies exchanges are not controlled by a central authority. It uses all the concepts of blockchain technology. Hence it does not hold anyone money or any private information.</li><li>In decentralized cryptocurrency exchanges, you may find low trade volume since it is quite difficult to use and have limited functions. Most of the decentralized exchanges do have very little liquidity as compared with Centralised cryptocurrency exchanges.</li><li>Also, it allows only the exchange of cryptocurrencies. All the payments in decentralized cryptocurrencies exchanges are made using the cryptocurrencies.</li><li>Decentralized cryptocurrency exchanges are more secure since all the does are distributed and it becomes difficult to hack such exchange. It has a huge level of privacy since transactions are anonymous.</li></ul><h3>Which one should you choose? Centralized Cryptocurrency exchange or Decentralised Cryptocurrency exchange in the market.</h3><p>If we talk about choosing a Centralised or Decentralised, Decentralised would be better as it does have advantages over centralized Cryptocurrency exchange. Since centralized cryptocurrency exchange is more likely to get hacked. Also, the government does not control cryptocurrencies, hence most people do prefer decentralized issuance of money.</p><p>Hence many crypto enthusiasts are looking forward to using DEX or Decentralised cryptocurrency exchanges. It totally depends on you which cryptocurrency exchange you want to choose.</p><h3>What are the different factors one should consider while choosing a cryptocurrency exchange?</h3><p>If you want to use any cryptocurrency exchange, you should make sure that the following factors are in yours favour. As such factors would allow users to do transactions easily.</p><ul><li>An exchange should be available in your country.</li><li>It is really important that you choose an exchange which is available in your country and accessible worldwide. Due to huge financial regulations, government and financial institutions do heavily govern different cryptocurrencies. Let us take an example, if you use a cryptocurrency exchange based in India, it would only accept Indian currency for buying cryptocurrencies hence, you cant make transactions using any other currency like dollar or yen.</li><li>Trading fees should below.</li><li>It is really important that whatever exchange you choose should charge low or competitive trading fees. Since high trading fees would take a good amount of your deposit or withdraw. Such fees are applicable to different activities like buying and selling. For example, cryptocurrencies exchanges do charge 0.0005 BTC for transacting bitcoin from one cryptocurrency exchange to another. Many times several cryptocurrency exchanges do not disclose any hidden fees. Hence it is your duty to check all the hidden charges if added.</li></ul><h3>Several other features of cryptocurrency exchanges.</h3><ul><li>A cryptocurrency should be available.</li><li>Several times it so happens that users are not able to find their cryptocurrency in which they want to trade. Although most of the exchanges do have different major cryptocurrencies, like Bitcoin, Ethereum, lItecoin, and many more. If any user would like to invest in any lower altcoin, then one should check if it is available in exchange or not.</li><li>Reputation.</li><li>It is really important to check if your exchange has a good reputation in the cryptocurrency market. Since several cryptocurrency exchanges are coming up and most of them are a scam. Hence you need to make sure about the volume and make sure it is famous and used by a huge number.</li><li>Security.</li><li>A secure cryptocurrency exchange would always make sure, all user’s funds are secure. Also, such secure exchanges are less prone to any hack. It is really needed to make sure, the cryptocurrency exchange is secure.</li></ul><p>Several other factors do matter while choosing a cryptocurrency exchange such as liquidity and many more. It is really important to make sure whatever exchange you choose, such exchange should allow you to sell an asset with less fluctuation in the prices. Most of the time liquidity occurs when more buyers and sellers come into the market. In most cases, big and famous exchanges do have high selling prices.</p><h3>What is the major difference between a cryptocurrency exchange and cryptocurrency wallets?</h3><p>Cryptocurrency wallets and cryptocurrency exchanges do have several differences. Hence one should not consider the wallet an exchange. Cryptocurrency exchanges let users store and manage different crypto assets. if you want to use a cryptocurrency wallet, then you have full control over your cryptocurrency like bitcoin. But it is not the same in cryptocurrency exchange. In cryptocurrencies exchange, you don’t have control over your cryptocurrencies, and many times it refers to as the crypto exchange.</p><p>A cryptocurrency wallet does let users fully control cryptocurrencies whereas exchange services do not. A digital cryptocurrency wallet is accessible from anywhere. But a major disadvantage of using a cryptocurrency wallet is that it does need to be safe. If you lost your wallet you won’t be able to recover your digital currency. But it is not in the case of any cryptocurrency exchange since in cryptocurrency exchanges you don’t have control and access over your private keys. Even if you uninstall any exchange application from your phone, you can easily access it from anywhere and use cryptocurrencies.</p><h3>Different types of wallets.</h3><p>Cryptocurrencies are of different categories, cold storage wallets, hardware wallets, software wallets, and paper wallets. Hardware wallets are just like small hardware storage devices like pen drives or hard disks. Whereas a software wallet is available on your desktop, you need to install it. And paper wallets are just like paper.</p><h3>What are some of the best cryptocurrencies exchanges?</h3><p>As a beginner, it is really important for you to choose a cryptocurrency exchange that is secure and has all the features and characteristics which let you do flawless transactions. Some of the famous cryptocurrencies exchanges are the following.</p><ul><li><a href="https://coinbase.com/">Coinbase</a>.</li><li>Coinbase is the most trusted cryptocurrency exchange. Such a cryptocurrency exchange has regulation by the united states financial institution. It does include around $250k deposits including the FDIC insurance. Also, it is back by several famous and mainstream investment banks.</li><li><a href="https://binance.com/">Binance</a>.</li><li>Binance is the most used cryptocurrency exchange, it was founded by Changpeng “CZ” Zhao. he is a Chinese software developer who built several systems such as the Tokyo Stock Exchange. Also further he developed trading software for trade book by Bloomberks. It was based in China but due to certain regulations, it moved to japan. Further due to Japan’s FCA it moved to Malta in Europe.</li><li>Binance lets users exchange coins to coin cryptocurrencies for trading. Some of the biggest coin pairs are available on such a platform. It does let users do margin and derivative tradings.</li></ul><h3>Some other exchanges.</h3><ul><li><a href="https://kraken.com/">Kraken</a>.</li><li>Kraken is one of the oldest and biggest cryptocurrency exchange. In the beginning, it was lagging, but recently it improved allot. It is one of the most trustworthy exchanges in whole space. All traders can to trade in both cryptocurrencies and fiat currencies. Several large numbers of cryptocurrencies exchanges are available. Kraken has some of the essential features like all-day support, a huge range of dining options, and many more.</li><li><a href="https://bitmex.com/">Bitmex</a>.</li><li>Bitmex is a best cryptocurrency exchange. Recently it has gained huge potential and gains but at the same time several losses. According to most of the reports, around 95% of traders lose money. It allows us to do trading based leverage hence most people do leverage trading without any caution.</li><li><a href="https://challengly.com/">Challengly</a>.</li><li>It is another cryptocurrency exchange which is fast and anonymously. It has got huge transaction speed. Challengely does not let users to invest or trade any cryptocurrencies.</li></ul><h3>Several decentralized Cryptocurrency Exchange.</h3><p>Some of the famous decentralized exchanges are the following.</p><ul><li>1inch.exchange.</li><li>It is a decentralized exchange like several others such as Kybernetwork, Bancor, Radar Relay, and much more.</li><li>Airwap.</li><li>It is a peer to peer token train platform on ethereum. It does not charge any trading fees.</li></ul><p>Several other famous DEXs are available in the market.</p><h3>Conclusion.</h3><p>It totally depends on you which<a href="https://techiesphere.co/"> cryptocurrency</a> exchange would you like to choose. As a newbie, you should use coinbase or binance since it is much easier to use and you don’t need to care about losses like in BitMEX. If you are intermediate or professional and would like to choose a trading cryptocurrency exchange, BitMEX would be a perfect choice.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4095257fa3b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Blockchain is the most demanding skill in 2020?]]></title>
            <link>https://medium.com/@adityadhir97/why-blockchain-is-the-most-demanding-skill-in-2020-d3de2264ef10?source=rss-34c84bf846ce------2</link>
            <guid isPermaLink="false">https://medium.com/p/d3de2264ef10</guid>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[blogging]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Aditya Dhir]]></dc:creator>
            <pubDate>Wed, 03 Jun 2020 14:01:34 GMT</pubDate>
            <atom:updated>2020-06-03T14:01:34.773Z</atom:updated>
            <content:encoded><![CDATA[<h3>Introduction</h3><p>According to Linkedin’s latest survey <a href="https://techiesphere.co/">Blockchain</a> is a most demanding skill in 2020. Despite the pandemic, it still rose as there is a huge demand in industries due to the shift to the internet, and also, everyone wants to store data and share it securely.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*goKgbwMzB1-rnnk7" /></figure><p>Blockchain technology’s foundation held by a group of people or an individual named Satoshi Nakamoto. Satoshi Nakamoto brought up a great paper on <a href="https://techiesphere.co/">Bitcoin: A peer to peer Electronic cash system</a>. In his paper, he discussed a digital currency, which is decentralised, distributed and secure.</p><p>In today’s time blockchain technology has grown a lot and is the most demanding skill and technology for securing data. No industry, in today’s time, is not using blockchain technology, whether it is education, supply chain, real estate and many more. Blockchain technology promises the development of the best-secured system for transaction data in a distributed and decentralized way.</p><h3>Reasons why many industries are using blockchain technology in 2020.</h3><p>Blockchain technology is a distributed ledger of immutable data stored in a decentralised network. Blockchain network is a distributed and decentralised, which means, anyone from anywhere an access the data, provided that the person is a part of the blockchain network. And there is no central authority to control blockchain network, hence it is decentralised. So, All the data stored in the blockchain is cryptographically used and stored in a tree called Merkle tree. Due to its distributed and decentralised nature, many industries are looking forward to using this technology for their businesses and securing data. You name, any industry, it would be using blockchain technology. Some examples are Real Estate, education, media and many more.</p><p>Blockchain technology helps several industries to securely store and share data among several stakeholders. Also, blockchain brings more trust among the people as no single authority controls the translation or network. hence All the network nodes in a blockchain network have the liberty to verify transactions and manage the blockchain. Also, several nodes use various advanced concepts in order to validate the data and data is stored only when it is validated. Most of the industries do want to share data securely among several stakeholders.</p><h3>Basic use of blockchain technology in the education industry.</h3><p>Let us take a very basic example, suppose you have completed your education from a college and you apply for a job. The hiring team wants to verify all your documents and validate it. All of them need to follow a very difficult process of validation, they need to contact your university, find out about your education history and get it from your peers. Well, it is really a hectic task and it takes a huge time. But, on the other way, if your university issues the certificates on the blockchain network then anyone could verify it in very less time. Yes, that’s the power of blockchain technology. So, in the education industry, colleges would have a blockchain network, on which certificates can be issued to students and several hiring teams and validate from that blockchain network.</p><h3>Different Industries where you can work as a blockchain technologist in 2020.</h3><p>Several industries are looking for blockchain technology experts. You may find jobs in various industries like, real estate, media, supply chain, education, healthcare, legal industries and many more.</p><p>You don’t need to be technical to work in such industries. Several jobs are also available for non-technical people who want to start a career in blockchain technology. Some of the jobs are available in marketing, community, design and much more. Hence you don’t need have the technical knowledge to be part of blockchain revolution.</p><p>For techies, if you know several programming languages you can be a blockchain developer and learn new languages used for blockchain development. Blockchain developers’ jobs are different and they need to do several tasks such as developing business ideas, plans and much more.</p><h3>How should you begin learning Blockchain technology in 2020?</h3><p>Blockchain technology is very easy to understand if you have knowledge about cryptography and distributed systems. Blockchain technology needs a basic understanding of how hashing works and what is the Merkle tree. All these are some of the prerequisites to start learning about blockchain technology.</p><p>Once you gain information about all such concepts you should go deeper and start learning about how blockchain works and how it works. Blockchain technology is a vast concept to understand and somewhat technical. Also, If you want to go into the Blockchain development field, you should have basic understanding of several frontend technologies for testing UI for your blockchain application.</p><h3>Blockchain career map.</h3><p>If you want to learn about blockchain you can consider several choices below.</p><ul><li>A degree in Blockchain.</li><li>Some of the best universities in the world offer the best courses on blockchain development.</li><li>Moocs.</li><li>Nowadays, since Moocs have some of the best courses in the world to learn new things.</li><li>On the job training.</li><li>Several companies do provide on-field training if you get hired to work for a company as a blockchain developer.</li></ul><p>If you really want to be a blockchain developer you should have the hunger to learn about new things coming up every time. You can visit several forums from where you can learn blockchain development.</p><p>Also, You really need to be flexible while choosing a blockchain career. an individual can be a part-time coder, part-time entrepreneur or youtube celebrity. Hence, You just need good skills in all such fields.</p><p>You should have a good amount of knowledge about cryptocurrencies. Bitcoin is one such popular cryptocurrency and also several other cryptocurrencies are available.</p><h3>Some of the best Blockchain application platforms are following. You can start learning any of those.</h3><p>Ethereum is one such platform which you can use to learn about blockchain development and build some great decentralised application. Also, Ethereum lets you build both public and private blockchain.</p><p>For building applications on ethereum, you need to have knowledge about solidity programming language. Since, It is a specific language smart contracts, for creating decentralised applications.</p><p>Hyperledger is one more platform for building blockchain applications. It lets users build private blockchain networks for their blockchain application. If you want to create a blockchain application for your business you must use hyperledger.</p><h3>Conclusion.</h3><p>Blockchain technology careers are going to stay for a long time. hence, People who haven’t heard about <a href="https://bitcoin.org/bitcoin.pdf">bitcoin</a>, are trying to learn such technology. Since Blockchain technology has great potential to revolutionise several industries and the most demanding skill in 2020. Several career opportunities are available in the blockchain world. Hence, If you want to start a career in blockchain, it is the right time to do so because blockchain is the most demanding skill in 2020.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d3de2264ef10" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>